Study on the Maturity of Tomato Fruit by Imaging Hyperspectral Imaging System FS20 Series
In this study, a non-destructive and rapid high-throughput tomato fruit phenotype acquisition platform was developed based on the FS20 series of Imaging Hyperspectral Cameras to detect morphological and structural traits, component content traits, and comprehensive phenotypic traits of tomato fruits, respectively. A hyperspectral imaging system was used to obtain spectral images of tomato fruits. Then the spectral photos were analyzed to obtain morphological and structural traits (longitudinal diameter, transverse diameter, fruit shape index, and weight), followed by measurements of tomato fruit color traits (L*, a*, b*, c*, b*, and a*/b*), and component content phenotypic traits (hardness, SsC, lycopene, titratable acid, soluble sugar, and VC values). First, tomato fruit images were taken by a hyperspectral imaging system, and the morphological and structural traits were obtained by extracting the transverse and longitudinal diameter information of the spectral images using image processing methods. Secondly, we propose the indexes and thresholds for the classification of ripeness, which are essential for determining the optimal harvesting period of tomato fruits and transporting, storing, and preserving the fruits. Finally, a partial least squares regression (PLSR) phenotype prediction model for tomato fruit color and component content trait phenotypes was constructed using a continuous projection algorithm (SPA) to select feature wavelengths.
Figure.1-4 Different color classes of tomato fruits (Huang Yuping et al., 2019)
Fig.1-5 Different ripening stages of tomato fruits (Jiang et al, 2021)
1. Tomato fruit ripeness classification
a. Traditional methods
Confirmation of tomato fruit maturity: a/b tomato fruit color definition variable (Ariasetal,2000). Calculate a*/b*, based on the a*/b* value classification of tomato fruits of different maturity.
ii. Threshold values.
Table 1-1 Description of tomato a*/b* at maturity
|Mature class||a*/b*||Surface color|
|Green ripening stage||<0||0%|
|Color conversion period||0-0.5||0-30%|
Fig.3-5 Scatter diagram of chromatic aberration at maturity
b. Hyperspectral methods
i. Tomato hyperspectral data acquisition and analysis (process as follows)
Figure.3-1 Experimental flow chart
ii. Feature wavelength selection (PCA)
1. Clustering effect after PCA (3 different varieties).
Figure.3-6 Principal component analysis (a) ‘Saint Laurent tomato (b)’ Alexis T147 ‘tomato (c) “Cadyali 1832 tomato (d) three varieties mixed
2. Model fitting accuracy and location of feature wavelengths after PCA.
Figure.3-7 SPA selection process of characteristic wavelengths (a) is used to determine the minimum RMSE of 11 characteristic wavelengths (b) and the distribution of I1 characteristic wavelengths marked by each square.
c. Problems with this part
Is the gold standard chosen appropriately? In this section, the authors used a/b as the gold standard for ripeness measurement, and a/b was calculated from the spectra. Is it appropriate to use this standard to judge fruit ripeness?
FigSpec® series imaging hyperspectral camera adopts high diffraction efficiency transmission grating spectral module and high sensitivity surface array camera, combined with built-in scanning imaging and auxiliary camera technology; solving the traditional hyperspectral camera requires external pushing and sweeping imaging mechanism and focus complex and other difficult operation problems. It can be directly integrated with a standard C interface imaging lens or microscope to acquire spectral images rapidly.
FigSpec® series imaging hyperspectral camera application areas.
Spectral analysis, mineral screening, material sorting, fruit and vegetable analysis, geological exploration, agricultural remote sensing, industrial inspection, unmanned airborne hyperspectral, imaging analysis, portable hyperspectral imaging analysis, visible hyperspectral imaging analysis, infrared hyperspectral imaging analysis, thermal infrared hyperspectral imaging analysis, black plastic sorting, metal manufacturing, color sorting, gas detection, flame analysis, agricultural vegetation type identification, waste recycling fruit quality analysis, microscopic hyperspectral analysis, agricultural hyperspectral, remote sensing hyperspectral, spectral imaging analysis, vegetation hyperspectral, aviation hyperspectral, hyperspectral anomaly detection, hyperspectral fluorescence analysis, microscopic hyperspectral imaging, feature hyperspectral analysis, indoor hyperspectral analysis, criminal investigation hyperspectral analysis, soil hyperspectral analysis, environmental monitoring.